Krea compares Krea 2 with Google's Nano Banana 2
Krea has published a five-prompt comparison of its in-house Krea 2 image model and Google's Nano Banana 2, arguing that creators should keep both open instead of picking one.
Krea published a side-by-side comparison of Krea 2, its own foundation image model, and Nano Banana 2, Google's new general-purpose image model. Both are available in Krea's Image tool, where users can switch between them on the same canvas. From there, results can go to Krea's editor, video, upscaler or restyle tools.
Krea calls Krea 2 a model "built from scratch around aesthetics, style control, and taste." The company says Krea 2 is the core of its creative-control features: style references, moodboards and LoRA training all plug into it, so teams can train a reusable house aesthetic. Krea describes Nano Banana 2 as having "uncanny photorealism and very literal instruction following," with particular strength in text on objects, plausible branding and photographic depth.
The company ran five prompts through both models: a fashion figure in a lake, a cantilevered living room, a whisky pour, a hummingbird, and an anamorphic portrait. Krea says it used the same prompt for each model, with no per-model tweaking or cherry-picking. It describes Krea 2's outputs as composed and art-directed, often in studio-like settings. In Krea's account, Nano Banana 2 tends to add context the prompt didn't ask for, such as placing a whisky glass in a bar or a portrait on a cobblestone street, which makes its images read more like commercial photography.
This is a qualitative showcase from a vendor that sells access to both models, not an independent benchmark. The post gives no quantitative metrics, pricing or latency figures. The available text cuts off partway through Krea's guidance on when to use Nano Banana 2.